Implement Tobyfit like insturment resolution methods for quantification

[Ronald Fowler]

The quantification aspect of mantid requires the ability to perform an instrument resolution calculation, as is done in the current Tobyfit code. This will be used in simulation of the expected scattering of a foreground model and in fitting that model to experimental data within the GenericFit algorithm. A base class which performs the simultation will be inherited by each foreground function. This will be used in conjunction with the user defined S(Q,w) to calculate the expected scattering.

The simulation needs access to run data and to models for the effects of the moderator, the chopper, instrument dimensions, etc. Each MDEvent needs to represent the total signal in a single detector pixel.

[Ronald Fowler]

Adding instrument resolution class and models

Adding first parts of Tobyfit style instrument resolution simulation function Re #4960 . Class SimulateResolution will be inherited by user simulation functions and invoked via GenericFit algorithm.

Changeset: a48d3886040303f20153285311d1b8826aac9374

[Ronald Fowler]

Adding instrument resolution class and models

Adding first parts of Tobyfit style instrument resolution simulation function Re #4960 . Class SimulateResolution will be inherited by user simulation functions and invoked via GenericFit algorithm.

Changeset: a48d3886040303f20153285311d1b8826aac9374

[Ronald Fowler]

Re #4960 Update to simulate resolution function

Extended the models of resolution effects to include most of the tobyfit features. Still need to add the matrices mapping from one space to another and the extraction of parameters from ExperimentInfo.

Changeset: f63d622cb7cf8aa720efc9ce5ac0ee2e3c19cc62

[Ronald Fowler]

Re #4960 correct comments in simulate resolution

Corrected a few comments and minor change to function call. Added a throw if SimulateResolution called without correct initialisation.

Changeset: bd9c9f0edf5675a66cdfda1b4b1f50b997ea004c

[Ronald Fowler]

Re #4960 Updates to Tobyfit resolution methods

Added to code to SimulateResolution and RunParam to implement most of the operations in Tobyfit. While the basic structure is in place and individual procedures pass some tests, it still omits important functionality such as getting correct detector information from the instrument and any attempt to read from log files. CobaltSpinWave example has been updated for these changes, but no significant tests have yet been written. The test has been re-enabled in the CMakefile.txt.

Changeset: 855c489e1566fe6b846affbfddfeb1c2883ae9ce

[Ronald Fowler]

Re #4960 Updates to Tobyfit resolution methods

Basic simulation of resolution function is now possible for a foreground function such as CobaltSpinWaveDSHO. However the input data, including chopper and moderator parameters and detector information is read from a RunParam object. Much of this data should be extracted from the Instrument, the Sample and log files, though may be stored in the RunParam objects. Simple tests are consistent with Tobyfit results. A bug exists in taking account of sample shape.

Changeset: 24c6b5135d2ce01c9eb37384e4592e5ec39b093c

[Ronald Fowler]

Re #4960 Updates to Tobyfit resolution methods

QuadEnBackgroundTest.h changed to not check for success in one case. Test OK on Linux but fails on Windows, Mac.

Changeset: 1183d67a20a54fbb3d2de7b3238f1e844f786d46

[Ronald Fowler]

Re #4960 Updates to Tobyfit resolution methods

Added to code to SimulateResolution and RunParam to implement most of the operations in Tobyfit. While the basic structure is in place and individual procedures pass some tests, it still omits important functionality such as getting correct detector information from the instrument and any attempt to read from log files. CobaltSpinWave example has been updated for these changes, but no significant tests have yet been written. The test has been re-enabled in the CMakefile.txt.

Changeset: 855c489e1566fe6b846affbfddfeb1c2883ae9ce

[Ronald Fowler]

Re #4960 Updates to Tobyfit resolution methods

Basic simulation of resolution function is now possible for a foreground function such as CobaltSpinWaveDSHO. However the input data, including chopper and moderator parameters and detector information is read from a RunParam object. Much of this data should be extracted from the Instrument, the Sample and log files, though may be stored in the RunParam objects. Simple tests are consistent with Tobyfit results. A bug exists in taking account of sample shape.

Changeset: 24c6b5135d2ce01c9eb37384e4592e5ec39b093c

[Ronald Fowler]

Re #4960 Updates to Tobyfit resolution methods

QuadEnBackgroundTest.h changed to not check for success in one case. Test OK on Linux but fails on Windows, Mac.

Changeset: 1183d67a20a54fbb3d2de7b3238f1e844f786d46

[Ronald Fowler]

Re #4960 Added an additional foreground model

Added Tobyfit model 111, Damped Hisenberg Spin Wave. This is model 111 in TF. Simple tests on one point are consistent with TF, without sample shape effects.

Changeset: b3803866f50c8b3fd7407826eda6e84c63d4c763

[Ronald Fowler]

Re #4960 Added an additional foreground model

Warning fix for above change.

Changeset: c2643b46ecf67fea574176348efcb0750305a073

[Ronald Fowler]

Basic simulation gives results that agree with Tobyfit for single test case on two foreground models. Actually extracting the correct information from Instrument, Sample and log files remains to be done, as does data fitting. These parts will be new tickets.

[Andrei Savici]

How is the MagneticFormFactor class different than the older (and more complete - has multiple js, not only j0. this allows better description of transition metal ions and rare earths) MagneticIon?

[Ronald Fowler]

Replying to Andrei Savici:

How is the MagneticFormFactor class different than the older (and more complete - has multiple js, not only j0. this allows better description of transition metal ions and rare earths) MagneticIon?

I am sorry, I never thought to check for an existing Magnetic form factor class in Mantid. As it appears more comprehensive then the Tobyfit version it certainly makes more sense to use it in place of the current one. I probably will not have time to alter this for the current release.

[Alex Buts]

interesting piece of code.

This class does everything what Mantid do with neutrons in one go.

Problems are obvious: only 10-component chopper instrument with fixed instrument components and implicit component models is implemented for indirect scattering only. Only user scattering model can be changed.

It would not be difficult to plug this block into an algorithm and use it as test bench for productivity/physical results comparison -- it should already work for the single case it was written and be very efficient.

Any modifications would request reimplementing it in terms of existing Mantid components and it is not only Magnetic ion but almost everything else from physical constants to units conversion

It is a pity it is not embedded in an algorithm -- then would be useful as a test-bench allowing to see resolution effects right now.

With this remark I would close this ticket for now.

[Stuart Campbell]

This ticket has been transferred to github issue ​5806